Georgia State University
ScholarWorks @ Georgia State University
Political Science Faculty Publications
Department of Political Science
6-2007
Iraq the Vote: Retrospective and Prospective
Foreign Policy Judgments on Candidate Choice
and Casualty Tolerance
Jason Reifler
Georgia State University, [email protected]
Christopher Gelpi
Duke University, [email protected]
Peter Feaver
Duke University, [email protected]
Follow this and additional works at: http://scholarworks.gsu.edu/political_science_facpub
Part of the Political Science Commons
Recommended Citation
Reifler, Jason; Gelpi, Christopher; and Feaver, Peter, "Iraq the Vote: Retrospective and Prospective Foreign Policy Judgments on
Candidate Choice and Casualty Tolerance" (2007). Political Science Faculty Publications. Paper 7.
http://scholarworks.gsu.edu/political_science_facpub/7
This Article is brought to you for free and open access by the Department of Political Science at ScholarWorks @ Georgia State University. It has been
accepted for inclusion in Political Science Faculty Publications by an authorized administrator of ScholarWorks @ Georgia State University. For more
information, please contact [email protected].
Polit Behav
DOI 10.1007/s11109-007-9029-6
ORIGINAL PAPER
Iraq the Vote: Retrospective and Prospective Foreign
Policy Judgments on Candidate Choice and Casualty
Tolerance
Christopher Gelpi Æ Jason Reifler Æ Peter Feaver
Ó Springer Science+Business Media, LLC 2007
Abstract In this article, we model the effect of foreign policy attitudes on
both vote choice and casualty tolerance, using survey data collected during the
2004 election. We show that prospective judgments of the likelihood of success in Iraq and retrospective judgments of whether the war in Iraq was right
are significant determinants of both vote choice and casualty tolerance. The
prospective judgment of success is key in predicting casualty tolerance, while
retrospective judgment of whether the war was right takes precedence in
determining vote choice. In addition, there is an important interaction between the two variables, so the effect of one is conditional on the value of the
other. We believe this is compelling evidence that foreign policy matters, and
that it matters in reasonable ways.
Keywords
Voting behavior Æ Casualty tolerance
Introduction
At first glance, the parallels between the 1992 and 2004 Presidential elections
appear striking. Both elections featured an incumbent named George Bush
who had enjoyed tremendous public support after launching a war in Iraq.
Despite the high levels of support, both President Bushes soon found
C. Gelpi Æ P. Feaver
Department of Political Science, Duke University, Durham, NC 27708, USA
J. Reifler (&)
Department of Political Science, Loyola University Chicago, 6430 N. Kenmore, Chicago, IL
60626, USA
e-mail: [email protected]
123
Polit Behav
themselves deadlocked in the polls against surprisingly robust Democratic
challengers, who were buoyed by widespread perceptions of a weak economy
and skyrocketing health care costs. Yet Bush ‘‘43’’ eventually won reelection,
while Bush ‘‘41’’ met defeat. What can explain this change in electoral fortunes? There are many possible and complementary explanations for why
Bush 43 won despite facing a daunting electoral environment. In this chapter,
we focus on one explanation that has been the subject of extensive public
comment: the role of the ongoing Iraq war.
Pundits reconcile the different fates by stating that ‘‘Commanders-in-Chief
do not lose elections in wartime.’’ For this law-like aphorism to hold,
incumbents must actually be able to obtain their party’s nomination. Harry
Truman and Lyndon Johnson were so unpopular—due in large part to public
disenchantment with American participation in the Korean and Vietnam
wars—they abandoned their efforts to run for reelection.1 Moreover, the
belief that Americans will not oust incumbent President’s during wartime flies
in the face of the even more well-entrenched view that Americans will not
tolerate casualties in war. The war in Iraq was barely a week old when
observers began to worry that news of combat fatalities would cause public
support to collapse (Elder & Nagourney, 2003, Purdum, 2003; Ricks, 2003). By
October of 2004, as Presidential elections loomed, more than 1,100 U.S.
military personnel had been killed in Iraq fighting against an insurgency that
continued more or less unabated. Had most pundits foreseen this turn of
events, they surely would have predicted a complete collapse of public support
for the Bush Administration and for the war in Iraq. Moreover, they would
have predicted a lopsided electoral defeat for the President in November.
With these competing views of how the war in Iraq would play out politically,
pundits and political observers had inadvertently pitted the irresistible force
(the effect of casualties on public opinion) against the immovable object
(an incumbent President during war).
The relationship between the war in Iraq, American casualties, and the
Presidential election is—to say the least—somewhat conflicted. In this article,
we seek to make sense of these contradictory expectations by providing a
more systematic and nuanced argument linking attitudes toward the war in
Iraq, tolerance for U.S. military casualties, and Presidential vote choice in
2004. We do so by uniting two distinct literatures, foreign policy voting and
support for using the military/casualty tolerance within a single theoretical
framework derived from Fiorina’s (1981) theory of retrospective voting. The
economic voting literature has long examined the relative weight of retrospective versus prospective evaluations in explaining presidential approval
and vote choice. The literature on support for the use of force and casualty
tolerance literature is also keenly interested in comparing the weight of retrospective and prospective evaluations. However, little work has tried to
1
Truman was eligible to run but withdrew after losing in the New Hampshire primary to Estes
Kefauver. Johnson announced his decision not to run for reelection after the Tet offensive in
Vietnam prompted challenges from Eugene McCarthy and Robert Kennedy.
123
Polit Behav
integrate these two literatures. This paper is an effort to examine the contribution of retrospective and prospective foreign policy evaluations to both
voting and casualty tolerance. Moreover, our analysis balances the importance
of normative judgments (retrospective evaluations of the ‘‘rightness’’ of the
war) against empirical judgments (prospective judgments of the war’s eventual success).
We argue that the willingness of the public to pay the costs of war and to
reelect incumbent Presidents during wartime are dependent on these two
attitudes and the interaction between them. In particular, we show that retrospective normative evaluations of whether President Bush ‘‘did the right
thing’’ in attacking Iraq and empirical prospective judgments about whether
the U.S. will ultimately be successful in Iraq are two critical attitudes for
understanding how foreign policy judgments affect vote choice and one’s
tolerance for casualties. Further, we show that the retrospective normative
judgments serve as a more powerful predictor for vote choice, while the
prospective empirical evaluations of mission success better predict continued
support for the war in Iraq. These claims are consistent with the broader
literature on how foreign policy influences voting behavior, and the literature
that examines the public’s response to war and casualties. We also show that
these retrospective and prospective judgments are interactive—a person’s
attitude on one conditions the effect of the other. This interaction operates on
‘‘political’’ support (vote choice) as well as ‘‘mission’’ support (casualty tolerance).
To our knowledge, no other work integrates political support for the
President and support for American war efforts into a single theoretical model
and uses the same predictor variables to explain the separate dependent
variable measures of vote choice and casualty tolerance. This article bridges
important gaps in the literature by using one theoretical framework to connect
two separate research questions: (1) whether and how foreign policy affects
political evaluation and choice, and (2) the conditions under which American
citizens will bear the financial and human cost of military missions. To us,
these two questions seem inherently linked—support for missions should
connect to the Commander-in-Chief responsible for executing such missions.
These judgments are connected, yet with important differences. We find that
the retrospective normative judgments are more important for vote choice,
and the prospective empirical judgments better explain casualty tolerance.
Literature Review
Foreign Policy and Political Behavior
Scholars have long been troubled by Americans’ inability to answer survey
questions ‘‘correctly.’’ Poor performance on surveys has led observers to view
the American public as an ill-informed lot with little ability to think coherently about the substance of politics (Campbell, Converse, Miller, & Stokes,
123
Polit Behav
1960; Converse 1964), with issues playing a small to non-existent role in
shaping citizens’ voting decisions. Foreign policy evaluations were viewed as
among the least likely to affect political choices. When forming attitudes
about the performance of the economy, citizens have their personal experience to fall back on. But foreign policy is so removed from the everyday lives
of most citizens, it was argued, that it is simply unreasonable to think that
what happened beyond U.S. borders would have a large impact on Americans’
political behavior. In support of this claim many studies showed, at best, weak
evidence that foreign affairs affected the voting decision (e.g. Almond, 1950;
Stokes, 1966). As Almond writes ‘‘Foreign policy attitudes among most
Americans lack intellectual structure and factual content.’’
Over the past two decades, however, two reinforcing lines of research have
suggested that foreign policy judgments may have a greater influence on
political behavior than previously thought. First, John Hurwitz and Mark
Peffley (Hurwitz & Peffley, 1987a, 1987b; Peffley & Hurwitz, 1993) demonstrate that citizens have reasonably structured attitudes concerning foreign
policy. Moreover, their research demonstrates that public attitudes toward
foreign policy affect political evaluations, and citizens respond in understandable ways to changing world events. Such evidence of a ‘‘rational public’’
regarding foreign affairs is widespread (Holsti, 1997; Shaprio & Page, 1988;
Wittkopf, 1990). Aggregate opinion on foreign and defense policy is
remarkably stable and changes ‘‘have seldom, if ever, occurred... without
reasonable causes, such as the actions of foreign friends or enemies or changes
in the United States’ position in the world’’ (Shaprio & Page, 1988, pp. 220–
221). Christopher Wlezien (1995, 1996) shows additional evidence that that
the public’s preferences for changes in defense spending respond well to actual levels of spending on defense, as well as to feelings about the Soviet
Union.
Second, not only does the public have well-structured and relatively stable
attitudes about foreign policy, but mounting evidence suggests that these
attitudes have an impact on political behavior. It has long been known that
economic evaluations have an effect on presidential approval and vote choice
(e.g. Kiewiet, 1983, Kinder & Kiewiet, 1979, 1981). An increasing amount of
evidence has emerged showing that foreign policy judgments matter as well as,
and in roughly equal magnitude to, economic evaluations. For example, in a
timeseries of aggregate quarterly presidential approval data, Nickelsburg and
Norpoth (2000) show that the President is as much ‘‘Commander-in-Chief’’ as
‘‘Chief Economist.’’ Adding major foreign policy events as predictor variables
to their model, these international events matter at least as much as economic
evaluations. Using individual level data from several national random sample
surveys conducted from 1983 to 1987, Wilcox and Allsop (1991) find approval
of Reagan’s foreign policy is consistently a good predictor of Reagan’s overall
approval, though its strength relative to domestic issues does depend on the
salience of economic or foreign policy issues. Taking one step further in the
causal chain linking attitudes and vote choice, Nincic and Hinkley (1992)
demonstrate that foreign policy attitudes influence candidate evaluations in
123
Polit Behav
the 1980 and 1984 presidential campaigns. Similarly, in an analysis of the 1980
and 1984 Presidential electoral choice, Aldrich, Sullivan, and Borgida (1989)
find that foreign policy issues were just as powerful a vote determinant as
domestic issues. The impact of foreign policy on electoral choice does appear
to wax and wane with the flow of current events. This pattern is hardly surprising. As we note below, survey responses regarding the nation’s ‘‘most
important problem’’ suggest that the economy is nearly always salient in the
minds of voters, while concern about foreign affairs varies substantially.
In sum, we point to three well-supported findings concerning foreign affairs
and political behavior: (1) Citizen attitudes about foreign policy are well
structured, (2) Foreign policy evaluations matter for presidential approval and
presidential vote choice, and (3) Citizen attitudes are generally stable, and
changes in attitudes reflect changes in the international arena.
While analysts generally agree that public opinion is stable and responds to
events, substantial differences of opinion exist concerning whether citizens
respond to the international events themselves, or if foreign policy attitudes
are mostly mediated by elite rhetoric and framing. Of course, American citizens are (mostly) incapable of witnessing international events without them
first being mediated by the press. Thus, the press certainly has the potential to
uniformly shape opinion—as long as the reporting from different bureaus is
substantially similar—regardless of how well the reporting reflects the reality
on the ground. If the press systematically reports international events differently than they occur, then we should expect citizen opinions to reflect media
coverage more than the ‘‘events themselves.’’ The Tet Offensive during the
Vietnam War may be a classic example of where the media were unified in
their reporting, but where the reporting arguably differed sufficiently from the
actual events. Thus the resulting changes in opinion were moved by the
reporting, rather than the event. More specifically, while Tet was a tactical
failure for the attacking Communist forces, it was reported to the American
public as a disaster for the U.S. and evidence of a stalemate (Johnson &
Tierney, 2006).
Even if the press accurately reports international events, political elites may
be able to control how the public understands the issues. Looking again to the
public’s preferences for changes in spending for defense, Witko (2003) finds
exactly this—the public does not respond to what the Soviet Union does, but
rather to how policy elites talk about the Soviet Union. This account is substantially in accord with Zaller’s (1992) account of mass opinion and attitude
change being driven primarily by elites and in many ways. Berinsky (2007)
goes one step further to argue that responses to international affairs are primarily endogenous to political predispositions.
Important questions remain that we hope scholars will address. First, to
what extent is press coverage of military conflict ‘‘accurate’’ and how would
we measure such accuracy? Second, how much latitude do elites have to
reframe press coverage of military conflict and how influential can their
framing efforts be? Third, to what extent are perceptions of international
affairs endogenous to political predispositions? A similar debate recently
123
Polit Behav
started taking shape on whether economic perceptions are endogenous (e.g.
Evans and Anderson 2006, Lewis-Beck 2006). While political predispositions
are likely to play a role in shaping individual level attitudes about international events (see Nyhan & Reifler, 2006; Reifler, nd for individual level
evidence concerning responding to information about international events in
the context of Iraqi WMD), it is also clear at the aggregate level that people
update their perceptions in reasonable ways. For example, over time the
percentage of the public who believe that Iraq had an active WMD program at
the time of the U.S invasion has greatly decreased.
Fortunately, for our purposes we do not need to resolve these debates in
order to proceed with our analysis. Our central focus is the impact of perceptions of the war in Iraq both on the willingness to continue fighting in Iraq
and the propensity to vote for Bush. We remain agnostic on the question of
whether these perceptions reflect ‘‘reality’’ or elite rhetoric, though as we note
in our conclusion, we think that our research focuses new attention on the
importance of addressing this debate. The endogeneity of foreign policy views,
on the other hand, is more of a concern and we return to this question in the
discussion.
Building on Earlier Research on War and Public Opinion
Ever since the Vietnam War, policymakers have worried that the American
public will support military operations only if the human costs of the war, as
measured in combat casualties, are minimal. Mueller (1973) argued that
public support for the Vietnam and Korean Wars dropped in proportion to
the log of casualties. In Mueller’s words: ‘‘While [the American public] did
weary of the [Korean and Vietnam] wars, they generally seem to have become
hardened to the wars’ costs: they are sensitive to relatively small losses in the
early stages, but only to large losses in later stages.’’ Casualties drained public
support, but only slowly.
This complex finding—that casualties have a more corrosive effect on
public support early in the war than they do later—gradually became simplified in the conventional wisdom to the view that the public will not tolerate
casualties. Edward Luttwak summarized the conventional wisdom well: ‘‘The
prospect of high casualties, which can rapidly undermine domestic support for
any military operations, is the key political constraint when decisions must be
made on which forces to deploy in a crisis, and at what levels’’ (Luttwak, 1996,
p. 36).
In other words, Mueller argued that the public was casualty sensitive. But
the conventional wisdom, pushed in part by missions like Somalia, translated
Mueller’s claim into a conviction that the public was casualty phobic (Hyde,
2000; Klarevas, 2000; Lane, 1998; Luttwak, 1994; Moskos, 1995; Record, 2000;
Sapolsky & Shapiro, 1996). Casualty sensitivity recognizes the human toll as a
cost of war; casualty phobia refers to a sensitivity so great that it amounts to an
unwillingness to support a military operation if even very low human costs are
incurred (Feaver & Gelpi, 2004). We hope to revise this conventional wisdom
123
Polit Behav
by demonstrating that the public’s willingness to pay the human costs of war
and its propensity to punish its leaders in the voting booth for using force vary
depending on specific attitudes about the war.
Casualty sensitivity is, to put it crudely, one’s price sensitivity to the human
cost of war. As with other forms of price sensitivity, some members of the
public are more sensitive to the costs than others. We see in the public a
continuum of casualty sensitivity ranging from the minimally sensitive—those
who view casualties as a necessary cost of war and not a determining factor in
shaping support—all the way to maximally sensitive, or those who support
only military missions that guarantee virtually no casualties.2 Over the past
decade, many scholars have worked to debunk the myth of a strictly casualty
phobic public.
While the view that the public is casualty phobic is widely entrenched
among policymakers and the elite (Destler & Kull, 1999), something close to
the opposite consensus has emerged in academic studies of American public
opinion. Bruce Jentleson (Jentleson & Britton, 1998; Jentleson, 1992) finds
that the public is ‘‘pretty prudent,’’ and will support paying even costly military operations provided the mission rationale conforms to certain standards.
Eric Larson (1996) argues that the public uses a rational cost-benefit model in
which the public depends very heavily on elite cues; when there is an elite
consensus (defined as congressional support) in favor of a military mission,
casualties are not highly corrosive to support. James Burk (1999) shows that
public support for missions did not collapse with casualties, even in the ‘‘hard’’
cases of Lebanon 1983 and Somalia 1993. Destler and Kull (1999) show that
public casualty tolerance even in ‘‘unpopular’’ missions like peace operations
in Bosnia is much greater than previously thought; moreover, public tolerance
of casualties is particularly a function of ‘‘international elite consensus’’ in the
form of multilateral support for the military operation.
Of course, even if the public as a whole is not casualty phobic, it still may be
the case that some people are. Feaver and Gelpi (2004) argue that there is
significant variation within the public regarding the use of force. They show
that public attitudes toward the use of force fit a quadripartite pattern: solid
hawks (roughly 30–35%) who will support virtually any military mission
regardless of the costs; solid doves (roughly10–30%) who will oppose almost
any mission regardless of cost; casualty-phobics (roughly 15–20%) who support a mission provided it is extremely low cost; and defeat-phobics (roughly
15–40%) who support a mission, despite mounting costs, provided that the
mission is likely to succeed, but who turn on a mission if it looks like it is
doomed to failure. The quadripartite pattern allows for variation in the proportion in each category, depending on the specifics of any given mission.
While many factors matter—stakes, elite consensus, type of mission, and so
forth—Feaver and Gelpi (2004) give pride of place to ‘‘expectations of
2
Some members of the public might also be insensitive to casualties because they are opposed to
the use of military force regardless of the number of U.S. casualties—that is some portion of the
public opposes the use of force even it results in no U.S. casualties. We account for this possibility
in our measurement of casualty tolerance.
123
Polit Behav
success.’’ When members of the public think victory is likely, they will support
paying the human costs of war. When they think victory is not likely, even
small costs will be corrosive of public support. In a comprehensive analysis of
aggregate public opinion trends from 1981 through 2004, Richard Eichenberg
comes to a similar conclusion: ‘‘successful military operations enjoy high
support, regardless of other factors that may be present’’ (Eichenberg, 2005,
p. 11).
Bringing the two literatures together—the role of foreign policy and the
impact of casualties on public opinion—produces the following baseline
expectation: mounting casualties should not by themselves jeopardize the
political fortunes of an incumbent, but the President does not have a blank
check. Instead, public support for the war and the president will depend upon
specific attitudes about the war. The public is able to reasonably incorporate
information from the international arena into its judgments, and these judgments affect presidential approval and vote choice. The public demands
success when its president goes to war, and it demands that the president
provide a good reason for the fighting. Neither of these factors—most especially perceptions of success—are entirely under the control of policy makers,
which brings us to the election of 2004.
Foreign Affairs as Salient Concern
If foreign policy judgments are going to matter in decision-making, then those
attitudes need to be accessible for the decision-maker (Aldrich et al. 1989).
Not surprisingly, foreign policy was a salient issue during the 2004 election
campaign. After all, during the four years that had elapsed since the 2000
Presidential election, the U.S. experienced a startling domestic attack from a
foreign enemy and the U.S. embarked on high-profile conflicts in Afghanistan
and Iraq. Looking at the Gallup poll’s most important problem over the last
several presidential elections, foreign affairs dramatically stands out as more
salient in 2004. The Gallup survey reports that in 2004 roughly the same
proportion of voters state international concerns (22%) as economic issues
(26%) as the ‘‘most important problem.’’ While 22% for foreign affairs is the
highest since 1984, the economy still trumped international concerns by a 2-to1 margin (47% to 23%) in Reagan’s reelection. In the three presidential
elections from 1992 to 2000, mentions of international affairs were 5% or less.3
Polling we conducted also shows that respondents report foreign affairs as
an important concern in the voting decision. In six separate surveys from
3
Gallup does not have apples-to-apples data for 1988. In most years, the ‘‘most important
problem’’ question allows respondents to mention more than one problem, i.e. the same
respondent could say both ‘‘the economy’’ and ‘‘foreign affairs.’’ Because of the multiple
responses, the cumulative total of the marginals exceeds 100%, sometimes by a wide margin. In
1988, Gallup used a single response format. In the single response format, 9% mention something
related to foreign affairs or defense policy as the most important problem. This proportion is still
more than double what was reported in the multiple response format during the 1990s.
123
Polit Behav
Table 1 Most important issue
3/5–3/18 3/19–4/2 4/3–4/16 4/17–4/29 6/18–6/28 10/21–11/1
Foreign policy issues like Iraq
and the War on Terrorism
Economic issues like jobs and
taxes
Social issues like abortion
and gay marriage
N
16%
19%
20%
24%
26%
30%
72%
70%
67%
63%
61%
59%
10%
9%
11%
11%
11%
10%
930
889
881
899
900
1,125
March 2004 to November 2004, we used a closed ended question to ask
respondents which issue (economy, foreign policy, social issues) was most
important to them personally when choosing which candidate to vote for.4
Because we only polled in this election, we cannot make inferences about the
importance of foreign affairs compared to other years based on this data
alone. However, Table 1 shows that the proportion stating foreign policy as
the most important issue doubled between the conclusion of the Democratic
primary campaigns and the general election in November.
In the 2004 election, then, foreign policy was a salient concern. The war in
Iraq was shaping up to be, in the words of Secretary of Defense Rumsfeld, a
‘‘long, hard slog’’—with a mounting human toll, making the election of 2004
an ideal place to examine the electoral politics of war.
The Model
We build on Fiorina’s (1981) model of retrospective voting to construct a
model that uses the same antecedent attitudes as predictors of vote choice and
casualty tolerance. Fiorina’s landmark work successfully synthesizes what
many saw as the irreconcilable traditions of behavioralism and rational choice
by creating a generalized voter’s calculus containing three distinct components: (1) political predispositions, (2) retrospective evaluations of the
incumbent, and (3) prospective judgments or future expectations.
We argue that two logically distinct attitudes—one’s willingness to continue
to pay a human cost in the war in Iraq and one’s vote choice in the 2004
election—are functions of retrospective and prospective judgments, as well as
one’s political predispositions (Party ID). Retrospectively, voters are judging
whether the decision to invade Iraq was the right one. Prospectively, voters
are judging whether the war in Iraq will turn out to be successful.
The impact of these attitudes on vote choice and casualty tolerance are
strongly intuitive. If the decision to invade Iraq was wrong, then it makes
4
The surveys were conducted by Knowledge Networks, which maintains a panel of respondents
recruited through Random Digit Dialing (RDD), who are equipped with WebTV and complete
surveys online. Detailed sampling information is available from the company website, http://
www.knowledgenetworks.com. Studies have found that the Knowledge Networks sampling
methodology yields representative samples (Couper, 2000; Krosnick & Chang, Unpublished
typescript), with results comparable to RDD telephone surveys.
123
Polit Behav
sense to elect a leader who would make (or would have made) different
choices. Thus we expect retrospective normative attitudes about whether
attacking Iraq was the ‘‘right thing’’ to matter most in determining vote
choice. If the war is hopeless, why continue to pay a price? Thus we expect
prospective attitudes about the likelihood of success to have the greatest
impact on casualty tolerance.
We expound on this intuition two steps further: we argue that (1) the
interaction of the retrospective (normative) and prospective (empirical)
judgments determines vote choice and casualty tolerance, and (2) the relative
weight assigned to retrospective or prospective judgments differs, depending
on whether it is vote choice or casualty tolerance that is in question.
We additionally argue that the effect of these two attitudes on vote choice
and casualty tolerance will be interactive. According to the logic outlined
above, we would expect attitudes about whether attacking Iraq was the ‘‘right
thing’’ to have little impact on casualty tolerance for respondents who feel
that success is unlikely. If victory is unlikely, the initial wisdom of the decision
to use force has little impact on the expected benefits (and thus the tolerable
costs) of the war. But if victory is likely, then attitudes about whether the war
was the ‘‘right thing’’ should have a substantial impact on the expected benefits from the conflict and influence casualty tolerance.
Similarly, the likely prospects for success should have little impact on one’s
judgment about the wisdom of using force if one does not believe that using
force was the ‘‘right thing’’ in the first place. However, if one believes that the
initial decision to use force was the ‘‘right thing’’ to do, then one’s attitude
about the likelihood of success should have a significant impact on one’s
judgment of the overall wisdom of U.S. policy, and thus on one’s willingness to
reelect the President.
Data and Methods
We propose here a latent variable approach in place of an explicitly spatial
model. Rather than utility functions of competing candidates, we see ‘‘Bush
support’’ and ‘‘casualty tolerance’’ as underlying attitudes expressed as latent
variables. Individuals possess an amount of ‘‘Bush support’’ or ‘‘casualty
tolerance’’, which we model as
yi ¼ x i b þ e i
ð1Þ
We keep this compatible with Fiorina (1981) by using independent variables
consistent with his generalized voter’s calculus. We include separate evaluations for political predispositions (long-term past experience), political past
experience (near-term experience under an incumbent), and future expectations, which yields the following equation:
yi ¼ PIDi þ RJi þ PJi þ ei
123
ð2Þ
Polit Behav
Table 2 Dependent variable marginals
Vote choice
Kerry
Lean Kerry
Undecided
Lean Bush
Bush
Casualty tolerance
40.5%
6.6%
2.2%
4.8%
45.9%
Oppose with 0 deaths
Support with 0 deaths
Support with 1,500 deaths
Support with 5,000 deaths
Support with 50,000 deaths
23.0%
23.7%
20.3%
17.2%
15.9%
In Eq. (2), yi* is how much ‘‘Bush support’’ or ‘‘casualty tolerance’’ one
possesses. The variables xi are expressed generally as the long term political
past experience (party ID or PID), political past experience under an
incumbent (retrospective judgments or RJ), and future expectations (prospective judgments or PJ).
Our Vote Choice and Casualty Tolerance dependent variables are both five
point ordinal scales. The data we analyze come from polling we conducted
immediately prior to the November 2004 election. Table 2 reports marginals
and full question wording is available in the appendix.
Because we are using ordinal scales to represent underlying latent attitudes,
ordered logit is an ideal estimation technique. We estimate the following
models for both vote choice and casualty tolerance:
yi ¼ b0 þ b1 Party ID þ b2 RightThing þ b3 Success
þ b23 RightThing Success þ ei
ð3Þ
yi ¼b0 þ b1 Party ID þ b2 RightThing þ b3 Success þ b23 RightThing Success
þ b4 Age þ b5 Education þ b6 Female þ b7 Minority þ ei
ð4Þ
Party ID is a standard partisan identification question, coded as Democrat
(–1), Independent (0), and Republican (1). RightThing and Success are both
four-point Likert scales. RightThing asks respondents if they approve of the
original decision to use military force against Iraq and is coded from ‘‘Strongly
Disapprove’’ (0) to ‘‘Strongly Approve’’ (3). Success asks respondents if they
think the U.S. is likely to succeed in Iraq, and is coded from ‘‘Not at all likely’’
(0) to ‘‘Very likely’’ (3). The independent variables are coded so that we
would expect to see positive coefficients in the ordered logistic regressions. In
other words, we expect a one unit increase in the independent variable to be
associated with a respondent possessing a greater quantity of the underlying
attitude under investigation, whether voting for Bush or casualty tolerance.
Table 3 shows the estimates from Eqs. (3) and (4) for both Vote Choice and
Casualty Tolerance.
As expected RightThing is a significant predictor of the both vote choice
and casualty tolerance-dependent variables, as is Success. Also as hypothesized, there is a significant interaction term.
123
Polit Behav
Table 3 Ordered logistic regression results
Vote choice
Model 1
Party ID
Iraq RightThing
Likelihood of Success
RightThing * Success
Model 2
1.60***
(0.12)
0.53**
(0.19)
–0.06
(0.19)
0.47***
(0.11)
Model 3
1.60***
(0.13)
0.54**
(0.20)
–0.03
(0.19)
0.46***
(0.11)
0.33
(0.17)
–0.31
(0.23)
0.04
(0.05)
–0.09
(0.09)
0.43
–641.49
972.05
1,007
Female
Minority
Age
Education
Pseudo R2
Log-Likelihood
Log-Likelihood v2
N
Casualty tolerance
0.43
–645.15
964.72
1,007
Model 4
0.33***
(0.08)
0.06
(0.13)
0.51***
(0.13)
0.29***
(0.06)
0.18
–1363.81
583.68
1,037
0.21*
(0.08)
0.12
(0.13)
0.66***
(0.13)
0.25***
(0.06)
–0.30*
(0.12)
–0.60***
(0.17)
0.11**
(0.04)
0.28***
(0.06)
0.19
–1333.51
644.28
1,037
* p < .05
** p < .01
*** p < .001
Because logit coefficients are extremely difficult to interpret directly, we
use predicted probability graphs to show the relationship between variables
and the importance of the interaction term. Figures 1 and 2 show predicted
probabilities of voting for George W. Bush (estimated from Model 2).
Consistent with our expectations (following from Fiorina) respondents’
.6
.4
0
.2
Probability
.8
1
Predicted Probability
of Voting for Bush
Strongly
Disapprove
Somewhat
Disapprove
Somewhat
Approve
Strongly
Approve
Iraq Was Right Thing
Likelihood of Success in Iraq
Very Likely
Somewhat Likely
Not Very Likely
Not Likely At All
Fig. 1 Vote choice predicted probabilities by RightThing (estimated from Model 2)
123
Polit Behav
.6
.4
0
.2
Probability
.8
1
Predicted Probability
of Voting for Bush
Not Likely
At All
Not Very
Likely
Somewhat
Likely
Very
Likely
Likelihood of Success
Iraq Right Thing
Strongly Approve
Somewhat Disapprove
Somewhat Approve
Strongly Disapprove
Fig. 2 Vote choice predicted probabilities by Success (estimated from Model 2)
retrospective judgments on whether attacking Iraq was the ‘‘right thing’’ has
the greater impact on vote choice. Figure 1 shows how the predicted probability of voting for Bush changes across values of RightThing, and the course
of this change depends on the values of Success. Although it can be hard to
discern directly from Table 3, the graph shows that RightThing always matters
in predicting the probability of voting for Bush. Certainly, RightThing matters
more as one ascends through Success from ‘‘Not at all likely’’ to ‘‘Very likely.’’
When we graph the predicted probability of voting for Bush as a function of
Success (Fig. 2), we find that this variable does not always matter. When
RightThing is ‘‘Strongly Disapprove,’’ Success has little to no effect on the
predicted probability of voting for Bush (this is the graphical way to interpret the
non-significant coefficient for Success in Table 3 in Models 1 & 2). The effect of
Success is significant for all other values of RightThing. When we examine the
predicted probability of voting for Bush across the values of Success, and see
how the change is affected by different values of RightThing, we see much more
modest differences in the change across Success, but much greater differences in
the predicted probability at the lowest point in the scale (‘‘Not at all likely’’).
Figures 3 and 4 show a nearly identical pattern, but in reverse. Consistent
with our expectations, when predicting whether one will tolerate at least 1,500
casualties in Iraq, Success becomes more important and RightThing explains
less of the action. When Success is ‘‘Not at all likely,’’ the probability of
tolerating 1,500 casualties is equally low regardless of the value of RightThing.
(Again, this is the graphical way to interpret the non-significant coefficient for
RightThing in Models 3 & 4.) Success, however, is always significant and is
responsible for a big change in the predicted probability of supporting the war
in Iraq, even if the U.S. suffers 1,500 casualties.
So what is the substantive significance of these attitudes and the interaction
between them? People who hold both beliefs—that the war was right and that
the U.S. will win—indicate the strongest support for continuing military action
even in the face of mounting casualties and for reelecting President Bush.
123
Polit Behav
.2
Probability
.4
.6
.8
1
Predicted Probability
of Tolerating at least 1500 Casualties
Strongly
Disapprove
Somewhat
Disapprove
Somewhat
Approve
Strongly
Approve
Iraq Was Right Thing
Likelihood of Success in Iraq
Very Likely
Somewhat Likely
Not Very Likely
Not Likely At All
Fig. 3 Casualty tolerance predicted probabilities by RightThing (estimated from Model 4)
.2
Probability
.4
.6
.8
1
Predicted Probability
of Tolerating at least 1500 Casualties
Not Likely
At All
Not Very
Likely
Somewhat
Likely
Very
Likely
Likelihood of Success
Iraq Right Thing
Strongly Approve
Somewhat Disapprove
Somewhat Approve
Strongly Disapprove
Fig. 4 Casualty tolerance predicted probabilities by Success (estimated from Model 4)
Likewise, people who hold the opposite view—that the war was wrong and
that the U.S. will lose—have the strongest opposition both to paying any more
human cost and to reelecting Bush. Think of the former as the ‘‘Bush Base’’
and the latter group as the ‘‘Vietnam Syndrome’’ crowd. The intermediate
attitudes—the ‘‘Noble Failure’’ view that the war was right but we will lose,
and the ‘‘Pottery Barn’’ view (you break it, you fix it) that the war was wrong
but we will win—operate in surprising ways.5 The Pottery Barn crowd is, on
average, more likely than the Noble Failure crowd to stomach continued
military action. In contrast, the Noble Failure subgroup indicates stronger
support for Bush.
5
We adopt the ‘‘Pottery Barn’’ label from Friedman.
123
Polit Behav
Table 4 Predicted probabilities
Bush Base (49%)
Noble Failure (8%)
Pottery Barn (15%)
Vietnam Syndrome (29%)
Vote Bush
Support 1,500+ casualties
83%
53%
26%
13%
89%
34%
40%
21%
What can these results tell us about why Bush 43 was able to win reelection
despite a costly and controversial war? The war did not help Bush 43 because
the public reflexively refuses to oust the Commander-in-Chief during wartime.
After all, a significant proportion of the public was strongly opposed to the
war and sought vigorously to remove the President as a result. Instead, the war
probably helped the President to win reelection and maintain support for the
war because he was able to persuade a majority of the public of two simple
points: (1) attacking Iraq was the ‘‘right thing’’ to do, and (2) the U.S. will
ultimately succeed in Iraq. Specifically, we found that 49% of our respondent’s
fell into the ‘‘Bush Base’’ category described above, while 15% were identified
as ‘‘Pottery Barn,’’ 8% as ‘‘Noble Failure,’’ and 29% as in the ‘‘Vietnam
Syndrome’’ category. Support from Bush Base and Noble Failure voters kept
the President in office, while Bush Base and Pottery Barn respondents
maintained popular support for the war (see Table 4).6 Of course, it is difficult
to say for certain whether the war ‘‘helped’’ Bush win reelection because the
appropriate counterfactual is not obvious.
To probe the robustness of our results, we re-run the analyses just among
those who say foreign policy is the most important issue and among those who
say economics is the most important issue. We find that the issue respondents
report as being most important slightly changes the relative strength of
our predictor variables, but that the overall causal story remains intact:
RightThing is a better predictor of the vote, while Success is a better predictor
of casualty tolerance (Table 5).
Restricting our analysis to just those respondents who cite economic concerns almost perfectly reproduces our findings (which should be expected
given their preponderance in the dataset). Goodness-of-fit statistics, such as
maximum likelihood analogs to R2, drop. But this should also be expected;
those who do not identify foreign policy as their primary concern ought to
have less variance explained with measures of foreign policy attitudes.
Replicating the ordered logits among the subset of respondents who report
foreign policy as the most important issue, we are able to explain a greater
proportion of the variance. If we are making a claim about the explanatory
power of foreign policy attitudes, it stands to reason that our model should
perform better among those who consider foreign policy the most important
6
The results in Table 4 are estimated from models not presented here (though available online).
In these models, we condense the four point scales in RightThing and Success to dummy variables,
and use the dummy variables as predictors. Results are substantively identical. We explain these
choices in more detail in the discussion section.
123
Polit Behav
Table 5 Logit models by most important issue
Dependent variable
Party ID
Iraq RightThing
Likelihood of Success
RightThing * Success
Female
Minority
Age
Education
Pseudo R2
Log-likelihood
Log-likelihood v2
N
Economy most important
Foreign policy most important
Vote
Casualty tolerance
Vote
Casualty tolerance
1.65***
(0.15)
0.70**
(0.25)
0.10
(0.25)
0.29*
(0.13)
0.27
(0.21)
–0.13
(0.27)
0.11
(0.07)
–0.09
(0.11)
0.36
–437.90
493.40
590
0.19
(0.11)
0.15
(0.17)
0.61***
(0.17)
0.21*
(0.09)
–0.58***
(0.15)
–0.49*
(0.21)
0.18***
(0.05)
0.25**
(0.08)
0.15
–788.01
283.41
605
1.63***
(0.35)
0.24
(0.52)
0.01
(0.43)
0.87**
(0.31)
0.24
(0.46)
–0.85
(0.62)
–0.25
(0.13)
–0.01
(0.24)
0.62
–104.22
333.07
305
0.35*
(0.17)
0.27
(0.27)
0.98***
(0.27)
0.12
(0.12)
–0.07
(0.21)
–0.96**
(0.34)
–0.00
(0.06)
0.19
(0.12)
0.24
–379.70
233.83
313
* p < .05
** p < .01
*** p < .001
issue. In explaining vote choice, the Pseudo R2 jumps. The interaction between RightThing and Success continues to be significant when we are predicting vote choice. However, when we examine casualty tolerance, the
interaction term loses its significance. Dropping the interaction term from the
model leaves both RightThing and Success as significant predictor variables,
with the latter having a coefficient more than two and a half times as large as
the former. While the interaction term is no longer significant, the relatively
greater importance of Success as a predictor variable conforms with our
overall story and explanation of how different attitudes affect vote choice and
casualty tolerance. The more one cares about foreign policy, the more one
focuses on the likelihood of success as the critical factor in deciding whether to
supporting the continuation of the war.
Discussion
While we are strongly confident in the strength of our results, we would be
remiss if we did not also point potential problems in our analyses. We see
three main threats to our inferences: (1) collinearity between our predictor
variables of theoretical interest, (2) an underspecified model, and (3) problems of endogeneity. Below we address each of these concerns.
123
Polit Behav
Multicollinearity
As one might expect, people who see the war in Iraq as the ‘‘right thing’’ and
also more apt to believe it will be successful. Similarly, those who think that
war was the wrong thing are less optimistic about success. Some may view the
variables as simply two different indicators of the same construct. If that is the
case, then our coefficients may not be well estimated, leading us to posit
differing causal significance to variables that are not actually distinct.
The Success and RightThing variables are highly correlated (rho = .66).
While high, there is still a substantial amount of variation between the two.
Table 6 reports the correlations between all of the independent variables we
use in the models above. As should be expected, Success and RightThing both
correlate strongly with the interaction made from those two variables. Table 7
presents variance inflation factor (VIF) scores for several different combinations of our independent variables. The VIFs are all below the standard ruleof-thumb threshold of 10 for excessive collinearty, except for the interaction
of the four-point Likert versions of Success and RightThing. Of course, the
interaction should exhibit more collinearity as it is a function of the two
Table 6 Correlations between variables
RightThing
RightThing
Success
RightThing
* Success
Female
Minority
Age
Education
Success
1
0.6599
0.9004
1
0.8303
–0.0810
–0.2269
–0.0253
–0.0226
–0.0453
–0.1177
–0.0560
–0.0416
RightThing
* Success
Female
Minority
Age
Education
1
0.0061
0.0009
–0.0329
1
–0.1609
–0.1453
1
0.0131
1
1
–0.0724
–0.2053
–0.0171
0.0002
Table 7 Variance inflation factor (VIF) values
Without
interaction
term
With
interaction
term
RightThing
Success
RightThing*Success
Female
Minority
Age
Education
RightThing (dummy)
Success (dummy)
RightThing*Success (dummy)
1.86
1.78
1.01
1.11
1.03
1.03
6.20
3.78
11.25
1.01
1.11
1.04
1.04
Mean VIF
1.30
3.63
Without interaction
term (dummy
variables)
With interaction
term (dummy
variables)
1.01
1.12
1.03
1.03
1.49
1.41
1.01
1.13
1.03
1.03
4.02
2.43
2.52
1.18
2.43
123
Polit Behav
variables. It is important to note that the predicted probabilities reported
above in Table 4 are calculated from a model using the dummy variable
versions for two reasons. First, the four named sub-populations are simply
easier to calculate with dummy variables. Second, we are reporting predicted
probabilities in which even the interaction term has a VIF below the standard
rule-of-thumb cutoff of 10.
Does the level of multi-collinearity threaten our inferences? We firmly
believe that it does not. Our core argument is that retrospective evaluations of
the rightness of the war better predict vote choice, whereas prospective
judgments of success better explain casualty tolerance. If we rerun the model
without the interaction, we find even stronger overall support for the claim
that Success is a more powerful predictor of casualty tolerance and that
RightThing is a better predictor of vote choice. As noted above, we obtain this
same result if we model our interaction with three dummy variables (which in
effect represent the ‘‘Bush Base,’’ ‘‘Noble Failure,’’ and ‘‘Pottery Barn’’
subgroups). This specification captures the interaction argument with a substantially lower level of multicollinearity—albeit without some theoretical
nuance—and yields the same empirical result. Thus the overall argument that
success trumps rightness when it comes to casualties is very strongly supported, as is the claim that rightness trumps success when it comes to voting.
Table 8 presents results of our models without an interaction term, and the
Table 8 Models without interaction terms
Vote choice
Model 1
Party ID
Iraq RightThing
Likelihood of Success
1.57***
(0.12)
1.30***
(0.10)
0.63***
(0.12)
Female
Minority
Age
Education
Pseudo R2
Log-likelihood
Log-likelihood v2
N
* p < .05
** p < .01
*** p < .001
123
0.42
–654.87
945.29
1,007
Casualty tolerance
Model 2
1.57***
(0.13)
1.29***
(0.10)
0.64***
(0.12)
0.32
(0.17)
–0.35
(0.22)
0.10
(0.08)
–0.08
(0.09)
0.42
–650.60
953.82
1,007
Model 3
0.35***
(0.08)
0.54***
(0.07)
0.94***
(0.09)
0.17
–1374.81
583.68
1,037
Model 4
0.21*
(0.08)
0.53***
(0.07)
1.03***
(0.09)
–0.31**
(0.12)
–0.65***
(0.17)
0.20***
(0.06)
0.30***
(0.06)
0.19
–1333.51
644.28
1,037
Polit Behav
results support the same substantive interpretation—RightThing matters more
for vote choice and Success matters more for casualty tolerance.
Under Specified Model
Another possible critique of the results presented here is that we have underspecified models. Certainly, there are many additional causes of vote
choice that one could include—ideology, opinion on important domestic issues, candidate traits, thermometer ratings, perhaps even spatial placement of
the competing candidates or other NES-type variables. We fully agree and
recognize that testing alternate specifications with these additional variables
could lead to stronger inferences. Unfortunately, we do not have those data
and are unable to test other model specifications. Our principal aim during
data collection was to test the public’s willingness to ‘‘pay’’ the cost of casualties as the ‘‘body bags started coming home.’’ As such, we elected to ask
fewer questions and draw additional samples at different points in time (and
therefore at increasing casualty numbers). Clearly, this is a difficult tradeoff.
We believe that we have made the correct choice since the question of the
public’s response to casualties and war in the context of an election is of
central concern to scholars and policy makers at this time.
While we did not ask about candidate placement or traits, we did ask
respondents about whether they believed that Bush or Kerry would do a
better job handling the economy and social issues. We retained our more
limited specification, however, because we were not sufficiently satisfied that
responses to questions about which candidate would handle issues better were
conceptually distinct from our vote choice question. Nonetheless, the results
in Table 3 remain robust even when we control for candidate preference in
handling the economy and social issues. Remember also that when we restrict
our basic models only to those who say that ‘‘foreign policy’’ was the most
important issue, model performance improves.
Another possible form of underspecification concerns the extension of our
results to other wars and elections. In Feaver, Gelpi, and Reifler (2005/2006),
we present aggregate level results that the importance of success in shaping
casualty tolerance is also present in other wars. We would acknowledge,
however, that the impact of wars on elections is likely to vary depending on
candidates’ positions on the war in question. Brody and Page (1972), for
example, find that attitudes toward Vietnam had little impact on electoral
choice in 1968 because Humphrey and Nixon differed little on this issue. Aldrich (1977), on the other hand, finds that Vietnam had a significant impact on
voting in 1972 because of the distinct positions taken by Nixon and McGovern.
The Iraq war in 2004 falls into the latter category because Bush and Kerry took
differing positions on whether the war was—in retrospect—the ‘‘right thing’’ to
do (Kerry’s initial vote for the war notwithstanding). If candidates do not take
different positions on the ‘‘rightness’’ of a war—or do not otherwise differentiate themselves along some other prominent war dimension—we would not
expect public attitudes toward the war to influence vote choice substantially.
123
Polit Behav
Endogeneity
Finally, one of the problems inherent in attitudinal research is that the causal
direction is unknown—the dependent variable may be influencing the independent variables. Choosing to vote for Bush or Kerry, for example, could
affect how one views the Iraq war. It is possible that the causal direction
actually begins with voting, and ends with evaluations of war (or other predictors in the model). In other work, we present additional analyses using
instrumental variables to try to isolate the causal influence of perceptions of
success and whether the war was the ‘‘right thing’’ to do. Moreover, we
conduct experiments that allow us to identify the causal impact of success on
casualty tolerance. We have not yet been able to conduct such experiments
regarding retrospective judgments of the ‘‘rightness’’ of the war and vote
choice, but we agree that such research is important and we hope to investigate this issue in the future.
It is worth noting, however, that we control for party identification in our
analyses, which we would view as causally prior to vote choice. Thus our
estimated coefficients already account for any relationship between partisanship and voters’ prospective and retrospective evaluations of Iraq. As a
result, if vote choice created the coefficients we observe for ‘‘success’’ and
‘‘right thing,’’ it must have been some aspect of vote choice beyond party
preference that did so.
Future Research
Perhaps the most pressing issue for future research is to develop a more
refined answer to the question ‘‘how do citizens judge success?’’ (and perhaps
even more importantly—’’are citizens using the correct metrics in evaluating
success?’’). To a large extent, the argument presented here is agnostic to
whether citizens make direct judgments about the war (or, at least as direct as
possible given that they are only exposed to what the media is able to report),
or whether citizens are mostly unaware of the ‘‘real’’ situation and simply
follow elite level debate. We tend to come down more on the side that citizens
follow what happens abroad via the news and can form judgments independent of simply parroting what their preferred elite says. But the results presented here are not affected by this prior question of where perceptions come
from.
Nonetheless, two logical next steps that flow from our research are: (1)
developing a better and more complete accounting of how citizens process
information about international affairs, and (2) developing a better understanding of how new pieces of information about foreign policy (including
how they are framed) affect judgments. We believe there is some exciting
work in this area, especially in the use of experiments (e.g. Berinsky, 2007;
Boettcher & Cobb, 2006; Tomz, 2007).
123
Polit Behav
Conclusion
The presidential election of 2004 was profoundly influenced by judgments
about the war in Iraq. But the public did not reelect the Commander-in-Chief
simply because there was a war on. Instead, the public appears to be drawing
carefully reasoned and reasonable judgments both about the war and about
the election. Our analysis indicates that a single theoretical model can be used
to describe both American’s willingness to support continued fighting and
their willingness to reelect the President. Specifically, we identify two attitudes—one prospective and the other retrospective—as key factors in shaping
opinion. Consistent with previous work on electoral behavior (Fiorina, 1981),
we find that retrospective judgments about the President’s decision to use
force are most influential in determining vote choice. Consistent with previous
work on casualty tolerance (Feaver & Gelpi, 2004), we find that prospective
judgments about the likelihood of success in Iraq are most important in
determining support for continuing to fight in Iraq.
Acknowledgments This work is supported by grants from the Carnegie Corporation and the
National Science Foundation. The authors would like to thank an anonymous reviewer, the editors of Political Behavior, John Aldrich, and seminar participants at Duke University for their
helpful comments.
Appendix
Most Important Issue: Respondents were asked ‘‘When choosing for whom to
vote, which of the following issues are most important to you personally?’’
Answer categories were ‘‘Foreign policy issues like Iraq and the War on
Terrorism,’’ ‘‘Economic issues like jobs and taxes,’’ and ‘‘Social issues like
abortion and gay marriage.’’
RightThing: Respondents were asked, ‘‘We would like to know whether you
think President Bush did the right thing by using military force against Iraq.
Would you say that you strongly approve, somewhat approve, some what
disapprove or strongly disapprove of his decision?’’ [Strongly approve = 3,
Somewhat approve = 2, Somewhat disapprove = 1, Strongly disapprove = 0].
Success: Respondents were asked, ‘‘Regardless of whether you think that the
President did the right thing, would you say that the U.S. is very likely succeed
in Iraq, somewhat likely to succeed, not very likely to succeed, or not at all
likely to succeed?’’ [Very likely to succeed = 3, Somewhat likely to succeed = 2, Not very likely to succeed = 1, Not at all likely to succeed = 0].
Vote: Respondents were asked ‘‘If the general election for President were held
today and the candidates were: George W. Bush, the Republican, John Kerry,
the Democrat, Ralph Nader, the Independent for whom would you vote?’’ If
undecided, respondents were asked ‘‘Toward whom do you lean?’’ [Bush = 4,
Lean Bush = 3, Undecided = 2, Lean Kerry = 1, Kerry = 0].
123
Polit Behav
Casualty Tolerance: To measure casualty tolerance, we asked respondents a
series of questions, including a split sample design to assess if the number of
dead and wounded affects opinion differently than just the number dead. We
found no difference by including information about the number wounded in
addition to those killed. In the questions that follow, the extra wording for the
split sample is included in parentheses.
All respondents received the statement, ‘‘Regardless of whether you think the
President made the right decision in attacking Iraq, as you know the United
States is engaged in an ongoing military operation there and has suffered
about 1,000 military deaths (and over 7,000 military wounded).’’
Respondents were then asked, ‘‘Would you support continued U.S. military
action in Iraq until a new Iraqi government can take over if it resulted in no
additional U.S. military deaths?’’ [If they said no Casualty Tolerance = 0, if
they said yes they were asked] ‘‘would you support continued U.S. military
action in Iraq until a new Iraqi government can take over if it resulted in up to
1,500 total U.S. military deaths (and over 10,000 military wounded)?’’ [If they
said no Casualty Tolerance = 1, if they said yes they were asked] ‘‘Would you
support continued U.S. military action in Iraq until a new Iraqi government
can take over if it resulted in up to 5,000 total U.S. military deaths (and over
30,000 military wounded)?’’ [If they said no Casualty Tolerance = 2, if they
said yes they were asked] ‘‘Would you support continued U.S. military action
in Iraq until a new Iraqi government can take over if it resulted in up to 50,000
total U.S. military deaths (and over 300,000 military wounded)?’’ [If they said
no Casualty Tolerance = 3, if they said yes Casualty Tolerance = 4].
Age: Seven category variable [18–24 = 0, 25–34 = 1, 35–44 = 2, 45–54 = 3, 55–
64 = 4, 65–74 = 5, 75+ = 6].
Education: Four category variable [High school or less = 0, High school diploma = 1, Some college = 2, Bachelors degree or higher = 3].
Female: Dummy variable [Male = 0, Female = 1].
Minority: Dummy variable [White Non-Hispanic = 0; Black, Hispanic or
other = 1].
References
Aldrich, J. (1977). Electoral choice in 1972: A test of some theorems of the spatial model of
electoral competition. Journal of Mathematical Sociology, 5, 215–237.
Aldrich, J., Sullivan, J., & Borgida, E. (1989). Foreign affairs and issue voting: Do presidential
candidates ‘waltz before a blind audience?’ American Political Science Review, 83(1), 123–141.
Almond, G. (1950). The American people and foreign policy. New York: Praeger.
Berinsky, A. (2007). Assuming the costs of war: Events, elites, and American public support for
military conflict. Journal of Politics, 69(4).
Boettcher, W., & Cobb, M. (2006). Echoes of Vietnam?: Casualty framing and public perceptions
of success and failure in Iraq. Journal of Conflict Resolution, 50(1), 831–854.
123
Polit Behav
Brody, R. A., & Page, B. I. (1972). Comment: The assessment of policy voting. American Political
Science Review, 66(2), 450–458.
Burk, J. (1999). Public support for peacekeeping in Lebanon and Somalia: Assessing the casualties
hypothesis. Political Science Quarterly, 114(1), 53–78.
Campbell, A., Converse, P. E., Miller, W. E., & Stokes, D. E. (1960). The American voter. New
York: Wiley.
Converse, P. E. (1964). The nature of belief systems in mass publics. In D. Apter (Ed.), Ideology
and discontent. New York: Free Press.
Couper, M. P. (2000). Web surveys: A review of issues and approaches. Public Opinion Quarterly,
64, 464–494.
Destler, I. M., & Kull, S. (1999). Misreading the public: The myth of a new isolationism. Washington, D.C.: Brookings Institution Press.
Eichenberg, R. C. (2005). Victory has many friends: The American public and the use of military
force, 1981–2004. International Security, 30(1), 140–177.
Elder, J., & Nagourney, A. (2003). A nation at war; POLL; Opinions begin to shift as public
weighs cost of war. New York Times March 26.
Evans, G., & Andersen, R. (2006). The political conditioning of economic perceptions. Journal of
Politics, 68(1), 194–207.
Feaver, P. D., & Gelpi, C. (2004). Choosing your battles: American civil-military relations and the
use of force. Princeton, NJ: Princeton University Press.
Fiorina, M. P. (1981). Retrospective voting in American national elections. New Haven: Yale
University Press.
Gelpi, C., Feaver, P. D., & Reifler, J. (2005/2006). Success Matters: Casualty Sensitivity and the
War in Iraq. International Security, 30(3), 7–46.
Holsti, O. R. (1997). Public opinion and American foreign policy. Ann Arbor, MI: University of
Michigan Press.
Hurwitz, J., & Peffley, M. (1987a). How are foreign policy attitudes structured? A hierarchical
model. American Political Science Review, 81(4), 1099–1120.
Hurwitz, J., & Peffley, M. (1987b). The means and ends of foreign policy as determinants of
presidential support. American Journal of Political Science, 31(2), 236–258.
Hyde, C. K. (2000). Casualty aversion: Implications for policymakers and senior military officers.
Aerospace Power Journal, 14(2), 17–27.
Jentleson, B. W. (1992). The pretty prudent public: Post-Vietnam American opinion on the use of
military force. International Studies Quarterly, 36(2), 49–74.
Jentleson, B., & Britton, R. L. (1998). Still pretty prudent. Journal of Conflict Resolution, 42(4).
Johnson, D., & Tierney, D. (2006). Failing to win: Perceptions of victory and defeat in international
politics. Cambridge, MA: Harvard University Press.
Kiewiet, D. R. (1983). Macroeconomics and micropolitics: The electoral effects of economic issues.
Chicago: University of Chicago Press.
Kinder, D., & Kiewiet, D. R. (1979). Economic grievances and political behavior: The role of
personal discontents and collective judgments in congressional voting. American Journal of
Political Science, 23, 495–527.
Kinder, D., & Kiewiet, D. R. (1981). Sociotropic politics: The American case. British Journal of
Political Science, 11, 129–161.
Klarevas, L. J. (2000). The polls-trends: The United States peace operation in Somalia. Public
Opinion Quarterly, 64(4), 523–540.
Lane C. (1998). Casualty attitude. New Republic, 219(17), 6.
Larson, E. (1996). Casualties and consensus: The historical role of casualties in domestic support for
U.S. military operations. Santa Monica, CA: Rand.
Lewis-Beck, M. (2006). Does economics still matter? Econometrics and the vote. Journal of
Politics, 68(1), 208–212.
Luttwak, E. (1994). Where are the great powers? Foreign Affairs, 73(4), 23–28.
Luttwak, E. (1996). A post-heroic military policy. Foreign Affairs, 75(4), 33–44.
Moskos, C. (1995). Grave decision: When Americans feel more at ease accepting the casualties of
war. Chicago Tribune December 12, p. 25.
Mueller, J. (1973). War, presidents, and public opinion. New York: John Wiley and Sons.
Nickelsburg, M. J., & Norpoth, H. (2000). Commander-in-chief or chief economist? The president
in the eye of the public. Electoral Studies, 19, 313–322.
123
Polit Behav
Nincic, M. & Hinkley, B. (1992). Foreign policy and the evaluation of presidential candidates.
Journal of Conflict Resolution, 35, 333–355.
Nyhan, B., & Reifler, J. (2006). Why are political misperceptions so difficult to correct? Paper
presented at American Political Science Association annual meeting, Philadelphia, PA,
September, 1, 2006.
Peffley, M., & Hurwitz, J. (1993). Models of attitude constraint in foreign affairs. Political
Behavior, 15(1), 61–90.
Purdum, T. S. (2003). A nation at war: The casualties; delicate calculus of casualties and public
opinion. New York Times March 27, B1.
Record, J. (2000). Force-protection: Fetishism sources, consequences, and (?) solutions. Aerospace
Power Journal, 14(2), 411.
Reifler, J. (nd). Causes and consequences of exposure to different news source. Unpublished
typescript.
Ricks, T. E. (2003). U.S. casualties expose risks, raise doubts about strategy. Washington Post
March 24, p. 1.
Sapolsky, H. M., & Shapiro, J. (1996). Casualties, technology, and America’s future wars.
Parameters, 26(2), 119–126.
Shapiro, R. Y., & Page, B. I. (1988). Foreign policy and the rational public. Journal of Conflict
Resolution, 32(2), 211–247.
Stokes, D. E. (1966). Some dynamic elements of contests for the presidency. American Political
Science Review, 60, 19–28.
Tomz, M. (2007). Domestic audience costs in international relations: An experimental approach.
International Organization, Forthcoming.
Wilcox, C., & Allsop, D. (1991). Economic and foreign policy as sources of Reagan support. The
Western Political Quarterly, 44(4), 941–958.
Witko, C. (2003). Cold war belligerence and U.S. public opinion toward defense spending.
American Politics Research, 31(4), 379–403.
Wittkopf, E. R. (1990). Faces of internationalism: Public opinion and American foreign policy.
Durham, N.C.: Duke University Press.
Wlezien, C. (1995). The Public as thermostat: Dynamics of preferences for spending. American
Journal of Political Science, 39(4), 981–1000.
Wlezien, C. (1996). Dynamics of representation: The case of US spending on defence. British
Journal of Political Science, 26(1), 81–103.
Zaller, J. (1992). The nature and origins of mass opinion. Cambridge: Cambridge University Press.
123